The proliferation of Internet of Things (IoT) devices has accelerated the demand for efficient, scalable, and secure data processing solutions. While cloud computing traditionally supported large-scale analytics, its limitations in latency, bandwidth, and privacy prompted the rise of edge computing as a decentralized alternative. This paper provides a comparative survey of edge computing architectures tailored for IoT environments, uniquely introducing a multi-layered taxonomy that encompasses far-edge, mid-edge, and near-edge sub-layers, alongside paradigms such as fog computing, cloudlet, and mobile edge computing (MEC). Unlike prior surveys that focused primarily on latency or deployment topologies, this study conducts a holistic analysis that simultaneously considers scalability, security, and sustainability trade-offs, supported by real-world use cases across healthcare, agriculture, manufacturing, and transportation. Intended for IoT system designers, network architects, and researchers in distributed computing, this survey highlights key architectural trade-offs and outlines future research directions for building optimized, intelligent edge-IoT systems.